Estimation of performance variance for networking congestion avoidance algorithm

The estimate of the variance for Transmission Control Protocol (TCP) Congestion Avoidance algorithm is derived. Congestion Avoidance algorithm provides stability of the data communication networks performing distributed control of the networking infrastructure (data links and routers). The performance of the protocol is basically defined by the relation between sliding window size (amount of data that a sender is allowed to send through the network without acknowledgement) and round trip time. The variance of this value is essentially important for network design and administration tasks. We consider piece-wise linear stochastic process of the sliding window size with the assumption that multiple data successfully delivered in sequence form a renewal process. Further analysis provides the estimate of sliding window expectation without applying Goelder’s inequality. This estimate is used then to derive the estimate of the variance. Conditions suitable for proper application of the estimate are obtained as well.